JOURNALOFNEUROTRAUMA28:2287–2306(November2011) ªMaryAnnLiebert,Inc. DOI:10.1089/neu.2011.1920 Comparison of Acute and Chronic Traumatic Brain Injury Using Semi-Automatic Multimodal Segmentation of MR Volumes Andrei Irimia,1 Micah C. Chambers,1,2 Jeffry R. Alger,3–5 Maria Filippou,6 Marcel W. Prastawa,7,8 Bo Wang,7,8 David A. Hovda,3 Guido Gerig,7,8 Arthur W. Toga,1,4,5 Ron Kikinis,9 Paul M. Vespa,6 and John D. Van Horn1 Abstract Althoughneuroimagingisessentialforpromptandpropermanagementoftraumaticbraininjury(TBI),thereis a regrettable and acute lack of robust methods for the visualization and assessment of TBI pathophysiology, especially for of the purpose of improving clinical outcome metrics. Until now, the application of automatic segmentationalgorithmstoTBIinaclinicalsettinghasremainedanelusivegoalbecauseexistingmethodshave, for the most part, been insufficiently robust to faithfully capture TBI-related changes in brain anatomy. This articleintroducesandillustratesthe combineduseofmultimodalTBI segmentation andtimepointcomparison using3DSlicer,awidely-usedsoftwareenvironmentwhoseTBIdataprocessingsolutionsareopenlyavailable. For three representative TBI cases, semi-automatic tissueclassification and 3D model generation are performed to perform intra-patient time point comparison of TBI using multimodal volumetrics and clinical atrophy measures. Identification and quantitative assessment of extra- and intra-cortical bleeding, lesions, edema, and diffuse axonal injury are demonstrated. The proposed tools allow cross-correlation of multimodal metrics from structural imaging (e.g., structural volume, atrophy measurements) with clinical outcome variables and other potentialfactorspredictiveofrecovery.Inaddition,theworkflowsdescribedaresuitableforTBIclinicalpractice and patient monitoring, particularly for assessing damage extent and for the measurement of neuroanatomical change over time. With knowledge of general location, extent, and degree of change, such metrics can be associated with clinical measures and subsequently used to suggest viable treatment options. Keywords:magneticresonanceimaging;segmentation;TBI;visualization Introduction neurocognitive function (De Beaumont et al., 2009). More- over,thelargenumberofrecentTBIcasesinsoldiersreturn- With an estimated 1.7 million people in the United ing from military conflicts has created a significant clinical States sustaining a traumatic brain injury (TBI) every challengeforUnitedStatesVeteransAdministrationhospitals year(Fauletal.,2010),themagnitudeofthismedicalconcern (Taber et al., 2006) and has highlighted the critical need for to the United States cannot be overstated (Chen and improvementinTBIcareandtreatment. D’Esposito, 2010). Every year, TBI cases are associated with Promptandproper managementof TBIsequelaecan sig- 1.2 million emergency room visits and over 50,000 deaths nificantlyaltertheircourse,reducemortalityandmorbidity, (Langloisetal.,2006).InterestinandpublicawarenessofTBI- reducelengthsofhospitalstay,anddecreasehealthcarecosts causeddiffuseaxonalinjury(DAI)hassurgedwiththepub- (Wattsetal.,2004).NeuroimagingofTBIisthereforevitalfor lication of studies suggesting that DAI in professional and surgical planning, by providing important information for amateurathletesmayhavebothacuteandchroniceffectson anatomical localization and surgical navigation, as well as 1Laboratory of Neuro Imaging, and 4Brain Research Institute, Department of Neurology, 2Henri Samueli School of Engineering and Applied Science, 3Department of Radiological Sciences, 5Division of Brain Mapping, Neuropsychiatric Institute, 6Brain Injury Research Center,DepartmentsofNeurosurgeryandNeurology,UniversityofCalifornia,LosAngeles,California. 7ScientificComputingandImagingInstitute,8SchoolofComputing,UniversityofUtah,SaltLakeCity,Utah. 9SurgicalPlanningLaboratory,DepartmentofRadiology,HarvardMedicalSchool,Boston,Massachusetts. 2287 2288 IRIMIAET AL. for guiding decisions regarding the aggressiveness of TBI 3Dbrainmodelgeneration,whilealsohighlightingtheadded treatment (Chesnut, 1998). Moreover, because TBI often re- clinicalinsightthatthisworkflowcanoffer. sultsincharacteristicimpairmentsdependentontheareaof involvement(Warneretal.,2010),lesionanalysiscanassistin Methods the localization of cognitive processes and in the attempt to Research participantsand MRIacquisition obtain novel information on the relationship between brain anatomy and behavior (Bates et al., 2003). For all these rea- Anonymizedneuroimagingdatafromthreerepresentative sons,reliableandprecisemethodsofTBIassessmentcanplay TBIsubjectswereacquiredattheBrainInjuryResearchCenter an essential role during both acute and chronic therapy for oftheUCLAGeffenSchoolofMedicine.Thestudyanddata thiscondition. acquisition protocol weredesigned andperformed in accor- Whereas neuroimaging is often critical for proper TBI dancewiththeDeclarationofHelsinkiandwereapprovedby clinical care, there is a regrettable and acute lack of robust theUCLAInstitutionalReviewBoard.Signedinformedcon- methods for the exploration, visualization, and quantitative sent was obtained from each patient or from their legally assessment of TBI-related anatomical insults and patho- authorized representative before any procedure was per- physiology. Starting 2–3 days after the acute injury, MRI is formed. In Table 1, the age, sex, and types of MR scans ac- generallyconsideredtobesuperiortoCTforthepurposeof quired from each patient are specified, in addition to the TBIassessmentandanalysis.WhereasCTisbetteratdetect- numberofdaysafterinjurywhentheacuteandchronicscan ing the pathology of bones and certain bleeds, MRI can sessions took place. In the case of the first two subjects, the identify bleeds more easily and more accurately as blood sequences being used are representative of MR acquisition composition gradually changes following TBI (Dubroff and protocols that are appropriate for patients requiring critical Newberg,2008).Inthiscontext,theapplicationofautomatic care. In the case of patient 3, we illustrate the use of an ex- MRIsegmentationalgorithmstotheclinicalinvestigationof tended MR scan protocol involving 12 distinct sequence TBI cases remains an elusive goal because many existing types. This protocol requires significantly more acquisition methods are insufficiently robust to accurately capture TBI- timeandcanbeenvisionedasapplicableforTBIpatientsin relatedchangesinbrainanatomy.Therefore,becausethestate highlystableconditionand/orwhocanwithstandremaining oftheart in quantitativeMRIanalysisofTBI ofteninvolves motionlessintheMRscannerfortheextendedamountoftime labor-intensive manual tissue classification, the clinician’s required for the acquisition of images. Both protocols dem- abilitytogeneraterobustandaccuratethree-dimensional(3D) onstrate the ability of our segmentation environment to ac- models of neuropathology for TBI diagnosis and treatment commodate,ontheonehand,theuseofalimitednumberof remains extremely limited. Consequently, despite recent scantypesand,ontheotherhand,theadditionalamountof progressinthedevelopmentofrobustimageanalysistools,it clinicallyrelevantinformationbeingsuppliedbyamoreen- remains difficult to quantify TBI-related brain insults either hancedprotocol.Throughoutthearticle,thesetof5sequences uni- or multi-modally, particularly for the purpose of im- that were used for the imaging of the first two patients is provingclinicaloutcomemetrics. referredtoasthe‘‘standard’’protocol,whereasthesetof12 To address the need for clinician-friendly TBI analysis sequences usedinthethirdpatientisreferredtoasthe‘‘ex- tools,wehereproposeandillustratethecombinedtheuseof tended’’protocol. multimodal,semi-automaticTBIanalysismethodswithin3D MRvolumeswereacquiredat3Tfromeachsubjectusinga Slicer, a freely available software environment for the seg- SiemensTrioTIMscanner(SiemensAG,Erlangen,Germany). mentation, registration, visualization, and quantification of ToassessthetimeevolutionofTBIbetweentheacuteandthe MRimages.Incontrastwithmanyothersegmentationenvi- chronicstage,scanningsessionswereheldbothseveraldays ronments for TBI that have appeared in the literature (Ding (acutebaseline)aswellasseveralmonths(chronicfollow-up) etal.,2008),the3DSlicerenvironmentisbothfreelyavailable after the traumatic injury event. To eliminate the effect of aswellaswidelyusedbyclinicians,scientists,andengineers. different scanner parameters during each scanning session, To showcase the ability to perform quantitative time point every subject was scanned using the same scanner for both comparison and assessment of TBI in 3D Slicer, we present acute and chronic time points. The MP-RAGE sequence threecasesofsemi-automaticTBIvolumesegmentationand (MuglerandBrookeman,1990)wasusedtoacquireT1-weighted Table1. Case Study SubjectData Specifying theAge,Sex, Cause of Injury, Dates of ScanningSessions as well asMR ModalitiesAcquiredfrom Each TBIStudy Case Dayof Dayof Causeof acute chronic T1pre- T2 GRE T1post- OEF DWI Patient Age Sex injury scan scan contrast TSE FLAIR T2 SWI Perfusion contrast EPSE DTI DWI1 2 mIP 1 45 M blunttrauma 2 176 O O O O O 2 31 M gunshot 2 201 O O O O O 3 33 M blunttrauma 3 238 O O O O O O O O O O O O ScanningsessiondatesarespecifiedasthenumberofdaysafteracuteinjurywhentheacuteandchronicMRvolumeswereacquired, respectively.CheckmarksindicateuseofeachMRscantype(seetextforclarification). TSE,turbospinecho;FLAIR,fluid-attenuatedinversionrecovery;GRE,gradient-recalledecho;SWI,susceptibility-weightedimaging;OEF, Oxygenextractionfactor;EP,echo-planar;SE,spinecho;DTI,diffusiontensorimaging;DWI,diffusionweightedimaging;mIP,maximum intensityprojection. COMPARISON OFACUTE ANDCHRONIC TBI 2289 images.Inaddition,MRdatawerealsoacquiredusingfluid- obtainedwasconfirmedusingGREimaging.TSET2-weighted attenuatedinversionrecovery(FLAIR)(DeCoeneetal.,1992), volumes were also used to confirm lesion characterization, T2-weighted turbo spin echo (TSE), also known as fast spin andDWIvolumeswereexamined,whereavailable,toaddi- echo(FSE),(Jonesetal.,1992),gradient-recalledecho(GRE) tionallydistinguishvasogenic(increaseddiffusion)fromcy- T2-weighted images, as well as susceptibility-weighted im- totoxic (restricted diffusion) edema (Huisman et al., 2003; aging(SWI)(Sehgaletal.,2005).Inthecaseoftheenhanced Warachetal.,1995).Non-hemorrhagicshearinglesionswere acquisition protocol for Patient 3, seven additional scan definedashyperintenselesionsthatwerevisibleprimarilyon sequences were also used, including diffusion weighted im- T2-weightedorFLAIRimages.Thismethodofidentification aging(DWI),diffusiontensorimaging(DTI),andmaximum was adopted because some shearing lesions often do not intensityprojection(mIP).Acompletelistofthesesequences show substantial amounts of hemorrhage but are primarily isprovidedforeachpatientinTable1,whereasTable2spe- visibleasareasofhyperintensityonFLAIR,T2-weighted,and cifiesstandardMRparametersassociatedwitheachsequence. GREimages. The appearance of hemorrhages in MR imaging is highly variable across sequences (Greenberg et al., 2009), as it de- Identification ofpathology pendsuponanumberofvariables,bothintrinsic(hemoglobin TBIcomprisesavarietyofcerebrallesiontypes,including oxygenation,erythrocyteintegrity)andextrinsic(e.g.,scanner hematoma, subarachnoid hemorrhage, contusion, and DAI. field strength, receiver bandwidth, T1 or T2 weighing). TBI edema,in particular, canfurther becharacterized byits BecauseSWItakesintoaccountMRsignalphaseinformation, extent and distribution ranging from perifocal (regional) to regionalmagneticfieldalterationscausedbyironconcentra- diffuse (generalized) (Unterberg et al., 2004). In this study, tiongradientsinthebrainareeasiertocapturewiththistech- non-hemorrhagic lesions on MRI scanning were coded as nique. As a result, SWI sequences are very useful in TBI hyperintensitiesonFLAIRimages.Byusingalonginversion imagingbecausesusceptibilityeffectsmakemicro-hemorrhages time(TI),theFLAIRsequenceachieveshighcontrastbetween moreobviousinSWIcomparedtootherparadigms.Because lesionsandhealthytissueaswellascerebrospinalfluid(CSF). SWIisthereforegenerallysuperiortoGREandT2-weighted Inthistypeofneuroimaging,lesionsappearashyperintense imagingindetectinghemorrhagiclesions(Tongetal.,2003), regions in the white matter (WM), surrounded by normal volumes acquired using this former modality were used to tissue of lower,moreuniform intensity (Itti et al.,2001). On identify micro-hemorrhages that were poorly detectable or theotherhand,T2-weightedGREimagingisexcellentforthe undetectable using other sequences. Specifically, following detectionofacutehemorrhagiclesions,andSWIismuchmore the previous guidelines of Tong and associates (2003), hem- sensitivethanconventionalT2-GREindetectinghemorrhagic orrhagic lesions were defined as hypointense foci that were DAI.Bycomparison,thesensitivityofDWIinidentifyingDAI not compatible with vascular, bone, orartifactual structures lesions is similar to that of FLAIR, and inferior to that of on conventional GRE images. Whenever foci etiology was T2-GREfordetectinghemorrhage(Huisman,2003).Although doubtful, the lesions in question were not considered to be GRE images do reveal some micro-hemorrhages, this se- hemorrhagic. quence is poor at resolving between gray matter (GM) and TBI volume segmentation is greatly improved with the lesions, on the one hand and WM and lesions, on the other availabilityofMRvolumesacquiredusingvarioussequence hand,atleastpartlybecauseGMlesionshavehigherrelaxa- typesthataresuitablefortheidentificationofpathology.As tion times than those of normal WM. For all these reasons, alreadypointedout,althoughFLAIRimagingissuitablefor brainlesionsadjacenttoCSFweresegmentedfromvolumes theidentificationofCSF-perfusedinjuries,itsabilitytodetect acquiredusingFLAIR,andthequalityofthesegmentations hemorrhage is greatly superseded in GRE T2 imaging and Table2. StandardMRAcquisition Parameters Associatedwith Each SequenceTypeSpecified in Table1 parameter TR TE TI FA ETL Thickness PhaseFOV Sampling Acquisition Sequence [ms] [ms] [ms] [deg] [mm] [%] [%] type Matrix pre-contrast T1 1900 3.52 900 9 1 1 100 100 3D 256·256 T2 TSE 3330 89 120 18 5 100 75 2D 512·512 FLAIR 8000 70 2375 130 16 3 75 75 2D 384·512 GRE T2 1500 7 20 1 3 75 80 2D 384·512 SWI 27 20 15 1 1.5 75 95 3D 192·256 Perfusion 2000 32 90 1 6 100 100 2D 128·128 post-contrast T1 1900 3.52 900 9 1 1 100 100 3D 256·256 OEFEP SE 10000 50 90 1 3 100 100 2D 128·128 DTI 8000 95 90 1 3 100 100 2D 128·128 DWI 1 4000 80 90 1 6 100 100 2D 130·130 DWI 2 4000 80 90 1 6 100 100 2D 130·130 mIP 27 20 15 1 12 75 95 3D 256·192 TR,repetitiontime;TE,echotime;TI,inversiontime;FA,flipangle;ETL,echotrainlength;FOV,fieldofview;TSE,turbospinecho;FLAIR, fluid-attenuatedinversionrecovery;GRE,gradientrecalledecho;SWI,susceptibility-weightedimaging;OEF,oxygenextractionfraction;EP, echo-planar;SE,spinecho;DTI,diffusiontensorimaging;DWI,diffusionweightedimaging;mIP,maximumintensityprojection. 2290 IRIMIAET AL. especiallyinSWI,becausethelatterhasimprovedabilitiesto detectmicro-bleeds.Inthepresentanalysis,itwasfoundthat non-hemorrhagic lesions were best delineated from FLAIR, and that the precise identification of hemorrhagic lesions requiredtheinclusionofGRET2andSWIimagechannelsin the segmentation. In the absence of these last two sequence types,thevolumeofnon-hemorrhagicpathologywasfound to inappropriately include that of hemorrhagic lesions, to theobviousdetrimentofsegmentationquality.Ontheother hand, inclusion of both GRE T2 and SWI but not of FLAIR causednon-hemorrhagic edematobemisclassifiedaseither healthy-looking GM or WM, which is reasonable to expect givenSWI’sexcellentsensitivitytothepresenceofironcom- pounds, although not to that of CSF perfusion of affected tissues.Finally,forthecasesexaminedhere,exclusionofSWI resultedintheinabilitytoidentifymicro-bleeds,whichpoints totheimportantroleofthissequencetypeinidentifyingthe locationsofsmallfocalinjuries. Image Processing Because TBI involves injury to both neuronal cell bodies andaxonalprocesses,globalatrophyofGMandWMareboth common in this condition. Partly for this reason, reliable segmentationofWMandGMlesionsisnecessaryforaccurate quantitative analysis of TBI lesions. Additionally, accurate quantificationoftotallesionvolume(lesionload)isessential for within-subject comparison of TBI volumes acquired at differenttimepoints.Toaddresstheseclinicalneeds,weused thesemi-automaticsegmentationtoolsavailablein3DSlicer software,includingtheAtlasBasedClassification(ABC)and ExpectationMaximization(EM)segmenters,whichhavebeen described elsewhere (Pohl et al., 2007; Prastawa and Gerig, 2008;Zoelleietal.,2007)andusedwidelytodetectpathology inMRscans(Prastawaetal.,2009).Briefly,theABCalgorithm can perform multimodal registration of MR images (Maes et al., 1997), tissue classification of the brain (Van Leemput et al., 1999),as well as lesion segmentation basedon outlier detection (Prastawa et al., 2003, 2004; Prastawa and Gerig, 2008).TheEMSegmenter,ontheotherhand,isanalgorithm guided by prior information represented algorithmically withinatreestructuretoestimateoptimalsegmentationviaa conventionalclassifier.Inthislattercase,thetreemirrorsthe hierarchy of anatomical structures and the sub-trees corre- spondtolimitedsegmentationproblems,asdescribedbyPohl andassociates(2007). As opposed to other specialized segmenters to which ac- cessisoftenrestrictedfromoutsideusers,bothABCandEM ‰ FIG. 1. Sample MR images acquired from Patient 1 using thestandardprotocol.Thefirstandsecondcolumnscontain images acquired at acute baseline and chronic follow-up, respectively.ShownfromtoptobottomareT1-weightedMP- RAGE images (A), T2-weighted TSE (B), fluid-attenuated in- version recovery (FLAIR) (C), T2-weighted gradient-recalled echo (GRE) (D), and susceptibility-weighted imaging (SWI)(E).Arrowsarecolorcodedtorepresentthefollowing: red, primary lesion(s); blue,smaller lesion(s); green, micro- hemorrhage(s);yellow,diffuseaxonalinjury(DAI).Notethat only representative (i.e., not all) injuries of each type are indicated by arrows. Color image is available online at www.liebertonline.com/neu COMPARISON OFACUTE ANDCHRONIC TBI 2291 segmenters are freely available as segmentation modules in was performed within 3D Slicer. Bias field correction was 3D Slicer. Both methods are automatic and their executions appliedusingafourth-orderpolynomialmodel.Becausethe requireminimalusersupervision.Inaddition,theABCseg- anatomyofsevereTBIoftendivergesfromhealthyanatomy menter possesses the ability to perform co-registration of to a very large extent, segmentation errors can frequently anarbitrarynumberofMRvolumesacquiredusingvarious occur, especially in severe TBI cases. To address this short- sequences. Thismakesithighlysuitabletothepresent mul- coming, manual review in 3D Slicer by an operator with timodalTBIimagingparadigms,whereasmanyas12distinct experienceintherecognitionofneurotraumawasperformed sequencetypesareused.Beforesegmentation,allimagevol- with the purpose of editing and correcting segmentation umes were co-registered to the pre-contrast MP-RAGE T1- errors. The guidelines of Filippi and associates (1998) for weightedvolumeacquiredduringtheacutebaselinescanning editingandsegmentationwerefollowed.Nevertheless,when session. Intensity normalization within and between scans assessing the extent of segmentation error, it is fair to note FIG.2. Three-dimensionalmodelsofTBIanatomyforPatient1(acutebaselinetimepoint),asgeneratedin3DSlicer.Eachrow displays one of six canonical views of the brain, whereas each column corresponds to a structure type (ventricles, non- hemorrhagic lesions, hemorrhagic lesions, and full model). In (A), ventricles are shown as opaque, whereas extracerebral injuries,graymatter(GM)andwhitematter(WM)volumesareshownastransparenttofacilitatevisualizationofeachstructure in relationship to cortical landmarks. In (B–D), ventricles are also shown as transparent for similar reasons. Arrows indicate TBI-relatedpathologyofinterest(seetextforelaboration).Colorimageisavailableonlineatwww.liebertonline.com/neu 2292 IRIMIAET AL. that,althoughoftenconsideredtobethegoldstandard,even to the volume at the latter time point. In addition to these manual outlining resulted in both moderate (3–10%) intra- measures, we also report five measures of atrophy, namely observer error as well as larger inter-operator variabilities thebifrontalindex(HahnandSchapiro,1976),thebicaudate (Raff et al., 2000). In conclusion, it is reasonable to expect index(Earnestetal.,1979),Evan’sindex(Syneketal.,1976), thepresenceofsubtlesegmentationinaccuracieseveninthe theventricularindex(Gyldensted,1977)andHuckman’sin- contextofourrigoroussegmentationparadigm. dex(Huckmanetal.,1975). Letthe bifrontal horndistancebethe maximumwidthof the anterior horns of the lateral ventricles (HLV). Then the Quantitative metrics bifrontalindexistheratioofthemaximumHLVwidthtothe Tissue-type segmentation was used to calculate the total inner skull diameter at HLV level. The bicaudate index is volumeofthreeselectedstructuretypes(ventricularsystem, definedasratiooftheminimumwidthofthelateralventricles non-hemorrhagiclesions,andhemorrhagiclesions).Thevol- (MLV) to the width of the inner skull at that level. Evan’s ume change was computed as the ratio of the difference in index is the ratio of the HLV to the maximum inner skull volumebetweenthefollow-upandacutebaselinetimepoints, diameter (MISD). Finally, the ventricular index is equal to FIG.3. AsinFigure2,forthechronicfollow-uptimepoint.Colorimageisavailableonlineatwww.liebertonline.com/neu COMPARISON OFACUTE ANDCHRONIC TBI 2293 MLV/HLVwhereasHuckman’sindexisthesumoftheMLV appreciatethe3DnatureofTBIaswellastheabilitiesof3D andHLVmeasures. Slicertodynamicallyrenderandcapturethecomplexitiesof severebrainpathology. Results Case1 Becauseneurotraumaoftenexhibitsmultiple,overlapping formsoffocaland/ordiffuseinjury,insightfromTBIimaging As Figure 1 reveals, Patient 1 exhibits a large deep-brain canbeenhancedthroughtheuseofmultimodaltechniques.In injurythat ishyperintense in theFLAIRimage(Fig.1C) be- whatfollows,theresultsofeachsegmentationand3Dmodel causeofperfusionbyCSFfromtheventricularsystem.Con- generationarereviewedandtheproceduresusedtoperform tralaterallywithrespecttothisinjury,asmallerinsultinthe tissueclassificationareillustrated.Foreachpatient,wepro- deepWMisalsoapparentinT1-weighted,FLAIRandGRE vide six canonical views of ventricles and non-hemorrhagic images(Fig.1A,C,andD,respectively).Theposteriorportion and hemorrhagic lesions, thus allowing the reader to fully oftheleftventricleseemslargerthanthecorrespondingpart FIG.4. TimepointcomparisonofTBIinPatient1using3DSlicer-generatedvolumesdisplayedinred(fortheacutebaseline time point), or green (for the chronic follow-up time point). As in Figures 2 and 3, each row displays one of six canonical viewsofthebrain,withwhitematter(WM)andgraymatter(GM)transparentinCandDtofacilitatevisualizationofeach structure inrelationship to cortical landmarks.Colorimageis availableonline at www.liebertonline.com/neu 2294 IRIMIAET AL. Table 3. Longitudinal Analysisof ThreeSelected Volumetric Structures(Ventricular System, Non-Hemorrhagic Lesions,and HemorrhagicLesions) inThree TBI Patients Volume [cm3] Structure Patient Acute Chronic D[%] Brain 1 1142.0 1009.3 -11.6 2 1098.5 978.7 -10.9 3 1111.2 1018.1 -8.4 Ventricular system 1 29.7 30.1 1.3 2 45.8 95.2 107.6 3 20.5 31.4 53.2 Non-Hemorrhagic lesions 1 84.7 19.9 -76.5 2 40.7 14.8 -63.6 3 54.7 0.9 -82.7 Hemorrhagiclesions 1 73.0 17.4 -76.2 2 51.3 <0.1 -99.7 3 22.5 0.3 -88.1 Values are reported for both acute and chronic segmented volumes, with the change D as a percentage being computed as (volume at the chronic time point—volume at the acute time point)·100/volumeattheacutetimepoint. of the right ventricle, possibly because of local WM inflam- mation. Additionally, FLAIR seems to reveal some DAI in the area surrounding the primary injury (yellow arrows), whereasGREdemonstratesthepresenceofhemorrhageboth adjacent to and remote from the primary insult (Fig. 1 D, greenarrows).Therefore,theinclusionofaGREsequenceis particularlyusefulinthiscasebecauseDAIcanbevisualized indirectlythroughshearhemorrhagescausedbybloodvessel lesions(Scheidetal.,2003).Figure1suggeststhattheabilityto characterize TBI is greatly enhanced by the use of SWI. As alreadyquantifiedindetailbyotherauthors(Tongetal.,2003, 2008) and illustrated in Figure 1E, this technique is very sensitive in the detection of extravascular blood products, which is illustrated by SWI’s identification of significant hemorrhage surrounding the primary injury, as well as dif- fuselythroughoutthebrain(Fig.1E).A3Danimationofthe segmentation for this patient is available in the Supplemen- tary Material (see online supplementary material at http:// www.liebertonline.com). InFigure2,theresultsoftheTBIsegmentationforPatient1 areshown.Eachrowdisplaysoneofsixcanonicalviews(left, right, dorsal, ventral, anterior, and posterior) of the MR- segmented brain, with every column corresponding to one particularstructuretype(ventricles,non-hemorrhagiclesions, hemorrhagiclesions,andthefullmodel).Ineachimage,ex- tracortical lesions, GM, and WM are transparent, whereas theventricularsystemisadditionallytransparentinrows2–5 tofacilitatevisualizingtherelativepositionofeachstructure type with respect to other cortical landmarks. Figure 2A shows 3D views of the ventricular system, with arrows drawing attention to anatomical changes caused by pathol- ogy.AspreviouslypositedbasedonFigure1,thisvolumetric analysisreflectsthelargeextentoftheprimaryinjury,aswell asthesignificantamountofbleedingpresentintheWM.The posterior part of the left ventricle is found to have been FIG. 5. As in Figure1, forPatient 2.Color imageisavail- slightlyenlargedasaconsequenceofthelesioncompressing able onlineat www.liebertonline.com/neu COMPARISON OFACUTE ANDCHRONIC TBI 2295 its anterior portion (Fig. 2A). By comparison, the posterior Figure 3 displays the segmentations of volumes acquired partoftherightventricleappearsthinner,althoughitsante- fromPatient1duringthechronicfollow-upsession.Although rior portion is larger because of the primary insult being theventricularsystemappearstoexhibitimprovedbilateral locatedcontralaterallywithrespecttothelatter(Fig.2Band symmetry (Fig. 3A) and the combined volume of bleeds as C). Figure 2B and C also illustrate the large volume of the extracted from GRE and SWI imaging appears to be much primaryinjury,whichcoversalargespatialextentmainlyin smaller than in theacute scan (Fig.3C), thelatter cannot be the left hemisphere and extends from the inferior to the su- statedregardingthevolumeofthescarcausedbytheprimary periorextremityofthebrain.InFigure2D,standardviewsof injury (Fig. 3B). In fact, the volume of the lesion does not the full 3D segmentation model are shown in which all appear to have decreased appreciably at follow-up. This is structuresarevisible,withtheexceptionoftheWMandGM confirmedinFigure4,inwhichavisualtimepointanalysis models,whichareomittedforvisualclarity.The3Dmodelof of the volumes associated with each structure is performed. subdural edema is shown as transparent (green color) so as Forthepurposeofthelatter,alllesionsintheacutebaseline not to obstruct the view of the brain. The full model per- volume (whether hemorrhagic or non-hemorrhagic) are dis- spective hastheadvantageofallowingonetovisualize and playedjointlyinred,whereasthelesionsandbleedsatfollow- summarize all forms of TBI-related pathology as displayed up are shown in green. This avoids the highly problematic intheothercolumns. task of differentiating lesions in the follow-up model FIG.6. AsinFigure2,forPatient2attheacutebaselinetimepoint.Colorimageisavailableonlineatwww.liebertonline.com/neu 2296 IRIMIAET AL. FIG. 7. As in Figure 2, for Patient 2 at the chronic follow-up time point. Color image is available online at www .liebertonline.com/neu according to their provenience (i.e., from hemorrhagic or thoseimagedatthechronicfollow-uptimepoint,whichagain non-hemorrhagicregionsintheacutebaselinemodel),while suggestssubstantialdecreaseininjuryextent.Thelongitudi- still allowing one to perform a visual and quantitative nalanalysisinFigure4suggeststhatPatient1exhibitsonlya comparisonofthetwotimepoints. small longitudinal change in the volume of the ventricular InFigure4,resultsofthetimepointcomparisonforPatient system. This is confirmed by the quantitative analysis re- 1aredisplayed.TheleftviewoftheGMvolumerevealsim- ported in Table 3, where it is seen that there is only a 1.3% portantdecreaseininflammationatthefollow-uptimepoint, increaseinventricularvolumebetweentheacuteandchronic particularlyoverdorsolateralfrontal,prefrontal,parietal,and time points. By contrast, in agreement with Figure 4, both superior temporal cortex. Comparison of the acute and hemorrhagic and non-hemorrhagic lesions are found to de- chronicWMvolumes,ontheotherhand,suggestssubstantial creaseinvolumesubstantially,thatis,byover76%.Among decrease in inflammation over the entire frontal and pre- the metrics used for quantification of atrophy, Evan’s index frontalcortex,aswellasmorediffuselyovertheothercortical registers the largest percentage change (11.42%) from the lobes. In this subject, comparison of the volumes associated acute to the chronic time point. (See Table 4 for numerical with lesions portrays the acute baseline lesions as encasing valuesofthisandothermeasuresineachsubject.)
Description: